Data-driven mechanistic analysis of digital therapeutic–assisted training and evaluation of personalized protocol effects in children with attention deficit/hyperactivity disorder

Introduction: Digital therapeutics (DTx) for children with ADHD have gained attention as adjunctive interventions to pharmacotherapy; however, evidence remains limited regarding the mechanisms underlying training performance decline and the clinical effectiveness of performance data-driven personalized protocols. This study aimed to identify the core mechanisms of performance decline using digital training performance data and to compare the effects of a mechanism-based personalized protocol with those of a standard protocol. Methods: A total of 40 children with ADHD (aged 5-12 years), diagnosed according to DSM-5 criteria, were randomly assigned to an experimental group (personalized protocol, n = 20) or a control group (standard protocol, n = 20). During week 1, all participants completed the same 10 training contents, from which accuracy, reaction time, and error types (commission and omission errors) were extracted to construct individualized performance profiles. During weeks 2-4, the experimental group received a personalized protocol in which content composition, difficulty, and feedback were adapted based on baseline performance profiles, whereas the control group followed a standard protocol. Total training volume was controlled to be identical between groups. Pre-post assessments were conducted using the Comprehensive Attention Test (CAT) sensitivity indices and the Korean ADHD Rating Scale (K-ARS). Baseline-adjusted ANCOVA and effect sizes were reported. Results: Performance data analyses indicated that performance decline was more closely associated with commission errors (impulsivity) than with omission errors, and a speed-accuracy trade-off was observed, whereby faster responses were associated with increased error rates. Discussion: These findings suggest that performance decline in digital training among children with ADHD may be related to impulsivity and speed-accuracy control characteristics. Furthermore, performance data-driven personalized DTx demonstrated greater improvements in objective attentional control and clinical symptoms than a standard protocol with an equivalent training volume. Clinical Trial Registration: https://cris.nih.go.kr, identifier KCT0010850.